Statistics of the Distribution of the Abundance of Molecules with

Subscriber access provided by UNIV OF MD - ALBIN KUHN LIB

Article

Statistics of the Distribution of the Abundance of Molecules with Various Drug Loads in Maytansinoid Antibody-Drug Conjugates Victor S Goldmacher, Godfrey Amphlett, Lintao Wang, and Alex Lazar Mol. Pharmaceutics, Just Accepted Manuscript • DOI: 10.1021/mp5007536 • Publication Date (Web): 30 Jan 2015 Downloaded from http://pubs.acs.org on February 2, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Molecular Pharmaceutics is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 22

Molecular Pharmaceutics

1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Statistics of the Distribution of the Abundance of Molecules with Various Drug Loads in Maytansinoid Antibody-Drug Conjugates

Victor S. Goldmacher1, Godfrey Amphlett, Lintao Wang, and Alexandru C. Lazar ImmunoGen, Inc., 830 Winter Street, Waltham, MA 02451, USA

1

Tel.: 1-781-895-0709. FAX: 1-781-895-0611. E-mail: [email protected]

ACS Paragon Plus Environment


Page 2 of 22

2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Table of Contents/Abstract Graphic:

ABSTRACT:

The maytansinoid antibody-drug conjugates (ADCs) in clinical development for cancer

therapy each contain a derivative of the microtubule-targeting agent, maytansine, covalently attached to the antibody via an engineered linker. A sample of any of these conjugate contains molecules with different numbers of maytansinoid molecules, or “drug” loads, the relative abundance of which can be determined by mass spectrometry. We examined the accuracy of the Poisson distribution and the binomial distribution in predicting the relative abundance of species with different drug loads for three antibodymaytansinoid conjugates with different antibodies and linker-maytansinoid pairings. We used variance, calculated from the experimental mass distribution data, as the parameter to determine the optimal value n of the binomial distribution number of trials. The accuracy of the Poisson distribution in predicting distribution of the species abundance in these conjugates varied among the conjugates. In contrast, the accuracy of the binomial distribution was similar for all three conjugates and comparable to the best accuracy of the Poisson distribution, as supported by a paired t-test.

KEYWORDS: Antibody-drug conjugate, ADC, maytansinoid, antibody, drug load, drug per antibody ratio, DAR, SMCC-DM1, SPP-DM1, sulfo-SPDB-DM4


Page 3 of 22


3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

INTRODUCTION A significant percentage of the antibody-drug conjugates (ADCs) in development today for the treatment of cancer contain a derivative of the microtubule-acting agent, maytansine, as the cytotoxic agent or “drug”.1 The maytansinoid is covalently attached to lysine residues of the antibody via a bifunctional linker, with different conjugates containing different linker-maytansinoid pairings. A typical human IgG antibody contains between 80 to 100 lysine residues, which have ε-amino groups, and an N-terminal amino group in each polypeptide chain, leading to random modification of these reactive antibody sites with linkers, and heterogeneity of the conjugates. The typical drug per antibody ratio (DAR) of an antibody-maytansinoid conjugate in the clinic averages between 3 and 4.1 As a result, each conjugate comprises a mixture of molecules that differ in the drug load (DL). (Here and below we use the term DAR as an average drug per antibody ratio among the molecules of the entire conjugate sample, and the term DL as the drug per antibody ratio of a given single molecule of the conjugate). Other approaches, including site-specific conjugation and/or other reactive sites on the antibody molecule are also being explored.2 It is of considerable theoretical and practical interest to find out if the distribution of the species with different DL in antibody-maytansinoid conjugates, as determined by examining their mass spectra, can be described by a statistical distribution model. Kim et al.3 conducted statistical analysis of conjugates of the humanized IgG1 antibody trastuzumab with the maytansinoid DM1 with various DARs. The authors computationally tested which of the five discrete statistical distributions, the geometrical, the negative binomial, the hypergeometric, the binomial, and the Poisson, would describe the mass spectrabased DL distributions most accurately. Only the binomial and the Poisson distributions were found to be satisfactory in matching the experimental data. Among the binomial distributions using various numbers n of the number of independent trials (interpreted as the number of different amino groups on antibody molecule available for conjugation) the distribution assuming n = 17 out of the total 88 lysine amino groups, gave the best fit. The peptide mapping analysis, however, detected lysines in at least 70 different locations that could be attached with the maytansinoid. Because of this apparent discrepancy, the authors



Page 4 of 22

4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

concluded that the Poisson distribution was more suitable for describing the mass spectra-derived relative abundances of various DL species in the trastuzumab-SMCC-DM1 conjugates, but noted that this conclusion might not be valid for other ADCs, which, in principle, might be better described by other statistical distribution models. Here we present statistical modeling of three antibody-maytansinoid conjugates, each made with a different monoclonal antibody and a different linker-maytansinoid pairing: SMCC-DM1, SPP-DM1, and sulfo-SPDB-DM4.

EXPERIMENTAL SECTION Materials. We prepared maytansinoid conjugates of three humanized IgG1 antibodies, denoted below as Ab1, Ab2, and Ab3. Ab1 was conjugated to the maytansinoid DM1 via the non-reducible linker SMCC. Ab2 was linked to the maytansinoid DM1 via SPP, and Ab3 to DM4 via sulfo-SPDB, both disulfide-containing linkers. These linkers, maytansinoids, and conjugation procedures have been previously described.4 The amounts of the linker and the maytansinoid were carefully controlled to produce a panel of conjugates with various DARs. Mass Spectrometry Distribution Profile Analyses. Maytansinoid distribution profile analyses of these conjugate samples were performed by size-exclusion HPLC (SEC) coupled with ESI-TOFMS (Waters LCT) as described previously. 5,6 Before analysis, each sample was first deglycosylated using endoglycosidase PNGase F (New England BioLabs); 250 units of enzyme were added to 100 µg of a conjugate, followed by incubation at 37oC overnight. The samples were loaded onto the SEC column (TSKgel SuperSW3000, 4.6 mm ID x 30 cm, 4 µm; TOSOH Bioscience) maintained at 60oC with an initial flow rate of 0.25 mL/min and eluted isocratically with mobile phase containing 50% acetonitrile (J. T. Baker) in water with 1% formic acid (EMD) and 0.02% trifluoroacetic acid (Thermo Scientific). During the elution of the conjugates, the column flow rate was reduced to 0.1 mL/min and directed toward the electrospray source of the mass spectrometer. The raw mass spectra were acquired from 5004500 m/z range and deconvoluted using MaxEnt1 computer algorithm, a part of MassLynx instrument operating software (Waters). The major peak areas were used to calculate , the relative abundance of


Page 5 of 22


5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

conjugate molecules with the DL = ( are non-negative integers) in the conjugate sample. Only major peaks were considered for DL calculations, while the corresponding satellite (minor) peaks were not taken into account in these calculation because the areas of the latter were always proportional to the areas of the respective major peaks. Peptide Mapping. The identification of the conjugation sites was performed by a peptide mapping method. Conjugate samples were desalted into ultra-pure water using an illustra NAP-5 Column (GE Healthcare Life Sciences), dried overnight in a centrifugal vacuum concentrator, reconstituted by adding 200 µL of denaturing buffer containing 6 M guanidine HCl (Sigma), 1.2 M Tris-HCl (Sigma) and 20 mM dithiothreitol (Pierce) at pH 7.7, and incubated at 37°C for 1 h. Thiol groups were alkylated by adding 12 µL of 1 M iodoacetic acid (Sigma), and incubating the mixture at room temperature in dark for 40 min. The alkylation reaction was then quenched by adding 7 µL of 1 M dithiothreitol in water. The alkylated sample was then run through an illustra NAP-5 column equilibrated with 5 mL of digestion buffer containing 50 mM Tris-HCl and 1 mM CaCl2 (Fluka) at pH 8.0. The sample was enzymatically digested with trypsin (Promega) for 2 h at 37oC using a protein-to-enzyme ratio of 1:25 (w/w), and stopped by adding 2.5 µL of trifluoroacetic acid. An aliquot of 20 µL of the digest was injected onto a BEH130 C18 column (1.7 µm, 2.1x150 mm; Waters) installed on a UPLC-QTOF system (Waters). Peptides were separated by gradient elution (mobile phase A: 0.05% TFA, 2% acetonitrile in water; mobile phase B: 0.05% TFA, 80% acetonitrile, 20% water). The UV and mass spectrometry data were acquired by MassLynx 4.1 (Waters). The raw LC/MS data were processed by Biopharmalynx (Waters) for peptide peak identification.

RESULTS Accuracy of the Poisson distribution in predicting the relative abundances of antibodymaytansinoid conjugate molecules of a given DL. In the previously published study, trastuzumabSMCC-DM1 conjugates of various DAR were prepared by different processes.3 In order to make antibody-maytansinoid conjugates Ab1-SMCC-DM1, Ab2-SPP-DM1, and Ab3-sulfo-SPDB-DM4 with



Page 6 of 22

6 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

differing DARs, samples were prepared using varied amounts of linker and maytansinoid in conjugation reactions. The mass distributions of the Ab1-SMCC-DM1, Ab2-SPP-DM1, and Ab3-sulfo-SPDB-DM1 were then examined using SEC-MS. Typical deconvoluted intact mass spectra of the deglycosylated conjugates are shown in Figure 1. Figure 1:

From these spectra we calculated the relative abundance of conjugate molecules with a given DL (SI Tables 1-3), and compared these distributions with the Poisson distribution, using the mean x̄ (DAR) calculated from the mass spectra:


Page 7 of 22


7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

x̄ =

(1)

where is the relative abundance of conjugate molecules with the DL = ( are non-negative integers) in the conjugate sample. In addition, we determined the DAR values by an independent method, UV spectrometry, and found for all three panels of conjugates that the DAR values determined by these two methods were in good agreement (SI Tables 1-3). The experimental relative abundance values in Ab1-SMCC-DM1 differed from those predicted by the Poisson distribution within -0.9% to 1.2% (SI Table 4). The paired t-test, performed for each triplet of experimental distribution and corresponding Poisson distribution, indicated that the differences between the experimental and related Poisson distribution values were mostly not significant i.e. within the 95% confidence interval (a representative t-test, for the conjugate with DAR 3.4, is shown in Table 1). Similar analyses were performed for the other two conjugates. The Poisson distribution predicted the experimental distribution of Ab1-SMCC-DM1, Ab2-SPP-DM1 Ab3-sulfo-SPDB-DM4 with decreasing accuracy. The experimental relative abundance values in Ab2-SPP-DM1 Ab3-sulfo-SPDB-DM4 differed from those predicted by the Poisson distribution between -2.1% to 3.0% and -4.8% and 5.4%, respectively. Accordingly, progressively larger fractions of the experimental and related Poisson distribution values differed significantly according to the t-test (Tables 2, 3). The decrease in accuracy of the Poisson distribution in predicting the experimental mass distributions within the three conjugates is illustrated in Figure 2. Figure 2:



Page 8 of 22

8 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

These results led us to search for another statistical model. A statistical model for the incidence of antibody-maytansinoid conjugate species varying in DL in which all modifiable amino groups are of an equal reactivity. Let’s make the following assumptions: (i) there are n amino groups on each antibody molecule that are modifiable (out of the total number of lysine ε-amino groups, plus the four N-terminal amino groups); (ii) the probability of a maytansinoid molecule being attached to any of these n amino groups does not depend on whether any of


Page 9 of 22


9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

the rest of the amino groups have been modified with a maytansinoid molecule; (iii) the outcome of each attempt of linking a maytansinoid molecule to an amino group results in only one of two mutually exclusive events; either maytansinoid is attached to it, or not; (iv) the probability p (0

Statistics of the Distribution of the Abundance of Molecules with

Recommend Documents